Download MetaTrader 5

How to make someone believe the backtest ? - page 2

To add comments, please log in or register
Icham Aidibe
10795
Icham Aidibe  
Sabil Yudifera Daeng Pattah:

Ok thanks I just want to be given many opinions.


I'm agree, self train is actually a must. I've myself implemented such an algo. Results aren't each month the same, but globally remains alike. Then only the sl/tp/trail to be optimized. Optimization is a very long process, particularly with MT4. 

More there.

How To Use Resilient Back Propagation To Train Neural Networks -- Visual Studio Magazine
How To Use Resilient Back Propagation To Train Neural Networks -- Visual Studio Magazine
  • visualstudiomagazine.com
It's more complex than back propagation, but Rprop has advantages in training speed and efficiency. Resilient back propagation (Rprop), an algorithm that can be used to train a neural network, is similar to the more common (regular) back-propagation. But it has two main advantages over back propagation: First, training with Rprop is often...
Mohammad Soubra
33378
Mohammad Soubra  
Sabil Yudifera Daeng Pattah:
Testing robots on live is very long and market conditions are always changing every month and year.

Does it mean that all the great known standard indicators have to be forgetten each month/year

The answer is no for sure..
But
i wonder
Need answer
Icham Aidibe
10795
Icham Aidibe  

Well, you can have a macd set via a backtest of 10 years. The duration of the test shows that probabilistically it's a good setting. But things are not fixed, markets are fluids, dynamics, influenced by politicals events (not only the economical calendar). 

Watch out crude oil :

 

There, your previously macd set via a 10 yrs backtest, won't match no longer and be sure that the oil market didn't change in a month.

Chunsheng Yang
179
Chunsheng Yang  
Sabil Yudifera Daeng Pattah:
I am a developer. Testing robots on live is very long and market conditions are always changing every month and year.
How i solved this problem ?

What you said,  "market conditions are always changing..."  is the truth, which is essential to solve your question.   Based on this truth, even live-test should not be "made-believable".


Sorry, only less than 1% will win in trading, a  zero-sum game, so I will not tell you the right answer that I have found .  


Yet, with due respect, I can just give you a direction: (almost) none of the answers here are from the winner niche.

(FYI, Self-trained approach is close to or right in the winner camp)



Chunsheng Yang
179
Chunsheng Yang  
Marco vd Heijden:

I do this all the time.

I mean the future thing... not the backtesting, that's a waste of time... you can't change the past, but you can change the future



This is merely using statistic tools (with your pic. showing "ANACONDA").  Many people here do this routinely; this is nothing extraordinary--"change the future" (with R or Python...)  is more than a overstatement :-)  
Marco vd Heijden
Moderator
6137
Marco vd Heijden  
Chunsheng Yang:
This is merely using statistic tools (with your pic. showing "ANACONDA").  Many people here do this routinely; this is nothing extraordinary--"change the future" (with R or Python...)  is more than a overstatement :-)  

Actually it is generated by Prophet and not Anaconda.

Anaconda is merely the framework for the data frames, and i do not believe you that many people use this, and i also do not believe you that markets change.

Most people just don't know what they are talking about they just talk and talk and talk but they have no idea what they are looking at.

And then others just blindly agree with it without thinking at all, like sheep.

Sabil Yudifera
27218
Sabil Yudifera  
Chunsheng Yang:
This is merely using statistic tools (with your pic. showing "ANACONDA").  Many people here do this routinely; this is nothing extraordinary--"change the future" (with R or Python...)  is more than a overstatement :-)  
Agree
Icham Aidibe
10795
Icham Aidibe  

Okay. But markets do change. None of that would have been predictable on the basis of economical calendar nor previous datas (backtested or trained only once).

Self adapting systems are a must if you feel not reoptimizing all the time - on that point I'm agree to be the sheep ;-)

Sabil Yudifera
27218
Sabil Yudifera  

Why so many different opinions :)

Icham Aidibe
10795
Icham Aidibe  
Chunsheng Yang:
...

Is it supposed to be reassuring ? Not sure you noticed, but the included testers is the only way to test an EA made in MQL4/5. 

MT4 needs data to be imported, it's not brokers datas, but you can use the exact same spec' as your broker do. But the process is fastidiousness, just to set the spread to variable, the FXT should be hacked - and damn ... duration .... 1 year = 145h !

MT5 is supposed to do the whole job alone, downloading datas, setting & importing datas, using datas with the broker spec'. And it seems to be fully done, but indeed ... live results differs a lot - some say much more than would a MT4 backtest.

Dude, are you throwing that crude truth to our face, with no real solution ?

Share your secret, how do you validate your MT4/5 EA to live trading sessions ?!

To add comments, please log in or register